Supercharge Your Innovation With Domain-Expert AI Agents!

Method and device for image compression and decompression based on variational auto-encoder

A self-encoder, training image technology, applied in the field of artificial intelligence, can solve problems such as the inability to guarantee the fidelity of decoded images

Pending Publication Date: 2021-10-12
HUAWEI CLOUD COMPUTING TECH CO LTD
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The above method can achieve image compression in the low bit rate range, but since the decoded image is reconstructed by a generative model, the fidelity of the decoded image cannot be guaranteed

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Method and device for image compression and decompression based on variational auto-encoder
  • Method and device for image compression and decompression based on variational auto-encoder
  • Method and device for image compression and decompression based on variational auto-encoder

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0044] The technical solution in this application will be described below with reference to the accompanying drawings.

[0045] figure 1 It is a structural schematic diagram of a VAE provided by this application. The VAE 100 includes an encoder 110 , a quantization module 120 , a rate-distortion control module 130 , an arithmetic encoding module 140 , an arithmetic decoding module 150 and a decoder 160 .

[0046] The encoder 110 is used to process the input image, such as performing noise perturbation and feature extraction on the input image.

[0047] The quantization module 120 is used to quantize the image features output by the encoder 110 for subsequent processing.

[0048]The rate-distortion control module 130 is used to constrain the distribution of image features according to the probability prior information, so as to control the code rate. The distance between distributions is used to represent the gap between the actual feature distribution and the theoretical fe...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention provides a VAE training method. The method comprises the following steps: acquiring a training image; adding first noise into the training image to obtain a noise-added training image; and training a variational auto-encoder through the noise-added training image, wherein the variational auto-encoder is used for compressing and decompressing the noise-added training image. The addition of the first noise in the training image is equivalent to the disturbance of the pixel distribution of the training image, and the feature of the first noise is known, so that the VAE can carry out coding by using the global context information around the disturbed pixel, and the information utilization rate of the training image at a low code rate is improved, and therefore, the performance of the VAE can be improved by training the VAE through the noise-added training image, and the fidelity of a decoded image is improved. Besides, the robustness of the feature extraction function of the VAE can be improved by training the VAE by adopting the noise-added training image, and the fidelity of the decoded image can be improved when a code stream is transmitted by using a noise channel.

Description

technical field [0001] The present application relates to the field of artificial intelligence, in particular to a method and device for image compression and decompression based on a variational autoencoder. Background technique [0002] Image compression can reduce redundant information in image data, so image compression is of great significance for improving image storage efficiency and transmission efficiency. Traditional image compression methods such as joint photographic experts group (JPEG) have good compression effects in medium and high bit rate areas, but in low bit rate areas, the compression effect of traditional image compression methods is not ideal. [0003] A new image compression method is to compress the coded image through variational autoencoder (VAE), which mainly uses the convolutional network and the corresponding nonlinear transformation to extract image features, and directly extracts image features Arithmetic coding to achieve compression purpose...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): H04N19/42H04N19/44H04N19/136H04N19/124H04N19/147H04N19/91G06T9/00
CPCH04N19/42H04N19/44H04N19/136H04N19/124H04N19/147H04N19/91G06T9/00
Inventor 戴文睿程德李刚骆继祥熊红凯
Owner HUAWEI CLOUD COMPUTING TECH CO LTD
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com